2008-01-11 22:58:00

by Vineet Gupta

[permalink] [raw]
Subject: Usage semantics of atomic_set ( )

I'm trying to implement atomic ops for a CPU which has no inherent
support for Read-Modify-Write Ops. Instead of using a global spin lock
which protects all the atomic APIs, I want to use a spin lock per
instance of atomic_t. This works well when atomic_t is unitary and
statically initialized using ATOMIC_INIT (where I can reset the
spinlock_t as well). However if atomic_t var is embedded within a
bigger struct which is allocated dynamically how to I init the
embedded spin lock. atomic_set ( ) is the closest choice, however I
don't think it's current usage in kernel code qualifies it to be
"initializer only".

Doesn't that defeat the intended usage of atomic_t as a opaque type
which can be "effectively" used to hide other architecture specific
stuff.

Thanks,
Vineet


2008-01-11 23:16:49

by Roland Dreier

[permalink] [raw]
Subject: Re: Usage semantics of atomic_set ( )

> I'm trying to implement atomic ops for a CPU which has no inherent
> support for Read-Modify-Write Ops. Instead of using a global spin lock
> which protects all the atomic APIs, I want to use a spin lock per
> instance of atomic_t. This works well when atomic_t is unitary and
> statically initialized using ATOMIC_INIT (where I can reset the
> spinlock_t as well). However if atomic_t var is embedded within a
> bigger struct which is allocated dynamically how to I init the
> embedded spin lock. atomic_set ( ) is the closest choice, however I
> don't think it's current usage in kernel code qualifies it to be
> "initializer only".

A simple way to handle this might be to use a separate array of
spinlocks and hash each atomic_t to one entry in the array. You could
look in asm-parisc and arch/parisc to see an implementation of this
that is already in the kernel.

- R.

2008-01-12 01:37:55

by Chris Snook

[permalink] [raw]
Subject: Re: Usage semantics of atomic_set ( )

Vineet Gupta wrote:
> I'm trying to implement atomic ops for a CPU which has no inherent
> support for Read-Modify-Write Ops. Instead of using a global spin lock
> which protects all the atomic APIs, I want to use a spin lock per
> instance of atomic_t.

What operations are you using to implement spinlocks?

A few architectures use arrays of spinlocks to implement atomic_t. I believe
sparc and parisc are among them. Assuming your spinlock implementation is sound
and efficient, the same technique should work for you.

-- Chris

2008-01-12 18:37:13

by Vineet Gupta

[permalink] [raw]
Subject: Re: Usage semantics of atomic_set ( )

> What operations are you using to implement spinlocks?

The cpu provides atomic exchange instruction (atomic Read/write
semantics) which forms the back end of spin lock code.
__raw_spin_trylock( ) atomically swaps the lock memory with a reg (set
to 1)

Using a pool of spinlocks rather that original idea of spinlock per
atomic_t anyways seems to be a better idea.
I really appreciate the responses. Thanks a lot

-Vineet